[Mristudio-users] bug in DTIstudio 3.0 (beta) tensor calculation?

[susumu]

Hi Yi,

The outlier-rejection requires diffusion-weighted images more than 6. I
think the behavior of the Version 3 is related to this. 

Let me use an analogy of a simple linear fitting. When there are only two
data points and you need to find a best-matching line, you have only one
answer, which is connecting the two points. In this case, you are SOLVING
the equation. If you have more than 2 points, you do FITTING, not SOLVING.
When you do the fitting, you can use the outlier detection and rejection.
For example, if you have 10 points and one of them is completely off, you
can set some criteria for rejection and remove such a bad data point. If you
have only two points and one of them is corrupted, you can't do anything.
You just have to use the data as is.

For DTI, the 6-orientation is the required number of images and you SOLVE
the tensor equation if you have only 6 DWIs. When you have, say, 12
orientations, you can use the outlier detection and can reject up to
(theoretically) 6 images. 

In addition, we are going to release a better version of the outlier
rejection soon. So, at this point, I would suggest you not to use the
outlier rejection for your data.

Susumu

-----Original Message-----
From: mristudio-users-bounces at mristudio.org
[mailto:mristudio-users-bounces at mristudio.org] On Behalf Of Yi Jiang
Sent: Tuesday, December 23, 2008 1:29 PM
To: DTI Studio, ROI Editor, Landmarker Questions/Support
Subject: [Mristudio-users] bug in DTIstudio 3.0 (beta) tensor calculation?

Dear All,

I have a DTI dataset with 1 b0 and 6 DWIs, matrix size 512x256x256 each. I 
tried both the 64bit version 2.4.01 (Oct 2007) and 64 bits version 3.0.0 
(Sep 2008) DTIstudio to do DTI tensor calculation.

For the version 2.4.01, no matter I used automatic outlier rejection or not,

with all default settings (such as the background threshold is 10), I was 
able to obtain reasonable diffusion tensor images such as FA or colormaps.

For the version 3.0.0 Sep 2008, if I used the standard linear fitting 
method, the resulted images were all reasonable.

If I used automatic outlier pixel rejection, the resulted images (FA, 
colormap, eigenvalue etc.) were all zero.

If I used the automatic outlier image rejection when its error > 3% 
(default), the resulted images (FA, colormap, eigenvalue etc.) were all 
zero.

If I used the automatic outlier image rejection when its error > 99%, the 
resulted images (FA, colormap, eigenvalue etc.) were all zero.

The weird thing is I have use the default settings in the version 3.0.0 
(with automatic outlier image rejection when its error > 3%) to calculate 
diffusion tensors for some other datassets (1 b0 and 12 DWIs, matrix size 
512x256x256 each), the resulted images were all reasonable. Only for this 1 
b0 and 6 DWIs one (similar or better SNR than the other 12 direction DTI 
datasets), the results are not consistent.

Can anyone help? Thanks a lot!

Best,
Yi



_______________________________________________
Mristudio-users mailing list
Mristudio-users at mristudio.org
http://lists.mristudio.org/mailman/listinfo/mristudio-users